I think we can probably rule out OpenAI and equivalent cloud services allowing people to use their APIs to run phone scams. It's even worse PR than bots saying

I think we can probably rule out OpenAI and equivalent cloud services allowing people to use their APIs to run phone scams. It's even worse PR than bots saying racist things. And if they need to train their own model, you can get a lot of slaves and poor wannabes for the price of one competent NLP engineer, and the slaves and poor wannabes are less likely to decide they're the brains of the outfit and cut you out of the loop.